Learning A Highly Structured Motion Model for 3D Human Tracking
Published 1 January 2002
Tao Zhao, Tianshu Wang, Heung‐Yeung Shum
Citations32
Generate an AI Snapshot to get a quick, structured summary of this paper.
Study Snapshot
ObjectiveStudy objective
MethodsResearch methodology
PopulationPopulation studied
Sample sizeSample sizes
OutcomesStudy outcomes here
ResultsStudy results comes here
LimitationsResearch study limitations comes here
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
A structured representation of complex motion is used and a two-step unsupervised learning approach is proposed to recover the natural “primitives” from unsegmented 3D-motion captured sequences of complex human motion to recover high level structure.
Abstract
This paper presents our work on learning high level structure from human motion sequences, and its applications in human figure tracking. We use a structured representation ("primitives" and their transitions) of complex motion and propose a two-step unsupervised learning approach to recover the natural "primitives" from unsegmented 3D-motion captured sequences of complex human motion. The structure recovery is done under the MDL (minimum description length) paradigm. Then the learnt dynamic model of human motion is used in the CONDENSATION framework to successfully track human motion in a video sequence. Experimental results of ballet dancing sequences demonstrate that our approach works well. The learnt structure is also used to synthesize new video sequences.
Keywords
Computer ScienceEngineering
International Journal of Computer VisionCONDENSATION—Conditional Density Propagation for Visual Tracking
4,885 Citations1998Michael Isard, Andrew Blake
The Condensation algorithm uses “factored sampling”, previously applied to the interpretation of static images, in which the probability distribution of possible interpretations is represented by a randomly generated set.
Computer Vision and Image UnderstandingThe Visual Analysis of Human Movement: A Survey
1,767 Citations1999Dariu M. Gavrila
A number of promising applications are identified and an overview of recent developments in this domain is provided, including work on whole-body or hand motion and the various methodologies.
Computer Vision and Image UnderstandingA Survey of Computer Vision-Based Human Motion Capture
1,755 Citations2001Thomas B. Moeslund, Erik Granum
A comprehensive survey of computer vision-based human motion capture literature from the past two decades is presented, with a general overview based on a taxonomy of system functionalities, broken down into four processes: initialization, tracking, pose estimation, and recognition.
Articulated body motion capture by annealed particle filtering
860 Citations2002J. Deutscher, A. Blake +1 more
The principal contribution of the paper is the development of a modified particle filter for search in high dimensional configuration spaces that uses a continuation principle based on annealing to introduce the influence of narrow peaks in the fitness function, gradually.
3-D model-based tracking of humans in action: a multi-view approach
702 Citations1996Dariu M. Gavrila, L.S. Davis
A vision system for the 3-D model-based tracking of unconstrained human movement and initial tracking results from a large new Humans-in-Action database containing more than 2500 frames in each of four orthogonal views are presented.
Style machines
697 Citations2000Matthew Brand, Aaron Hertzmann
This work approaches the problem of stylistic motion synthesis by learning motion patterns from a highly varied set of motion capture sequences, and identifies common choreographic elements across sequences, the different styles in which each element is performed, and a small number of stylistic degrees of freedom.
Lecture notes in computer scienceStochastic Tracking of 3D Human Figures Using 2D Image Motion
694 Citations2000Hedvig Sidenbladh, Michael J. Black +1 more
A probabilistic method for tracking 3D articulated human figures in monocular image sequences that relies only on a frame-to-frame assumption of brightness constancy and hence is able to track people under changing viewpoints, in grayscale image sequences, and with complex unknown backgrounds.
Learning and recognizing human dynamics in video sequences
623 Citations2002Christoph Bregler
A probabilistic decomposition of human dynamics at multiple abstractions is described, and how to propagate hypotheses across space, time, and abstraction levels is shown.
Video textures
574 Citations2000Arno Schödl, Richard Szeliski +2 more
This paper presents techniques for analyzing a video clip to extract its structure, and for synthesizing a new, similar looking video of arbitrary length, and combines video textures with view morphing techniques to obtain 3D video textures.
Journal of Artificial Intelligence ResearchIdentifying Hierarchical Structure in Sequences: A linear-time algorithm
557 Citations1997Craig G. Nevill-Manning, Ian H. Witten
SEQUITUR is an algorithm that infers a hierarchical structure from a sequence of discrete symbols by replacing repeated phrases with a grammatical rule that generates the phrase, and continuing this process recursively, resulting in a hierarchical representation of the original sequence.
IEEE Transactions on Acoustics Speech and Signal ProcessingThe segmental K-means algorithm for estimating parameters of hidden Markov models
410 Citations1990B.-H. Juang, L. R. Rabiner
The authors discuss and document a parameter estimation algorithm for data sequence modeling involving hidden Markov models that uses the state-optimized joint likelihood for the observation data and the underlying Markovian state sequence as the objective function for estimation.
A mixed-state condensation tracker with automatic model-switching
324 Citations2002Michael Isard, Andrew Blake
A significant development of random sampling methods to allow automatic switching between multiple motion models as a natural extension of the tracking process is presented.
A multiple hypothesis approach to figure tracking
283 Citations2003Tat‐Jen Cham, James M. Rehg
A probabilistic multiple-hypothesis framework for tracking highly articulated objects where the probability density of the tracker state is represented as a set of modes with piecewise Gaussians characterizing the neighborhood around these modes.
Monocular tracking of the human arm in 3D
203 Citations2002Lídia Gonçalves, Enrico Bernardo +2 more
This work proposes to use a recursive estimator for arm position, and to provide the estimator with error signals obtained by comparing the projected estimated arm position with that of the actual arm in the image.
Impact of dynamic model learning on classification of human motion
90 Citations2002Vladimir Pavlović, James M. Rehg
A novel approximate structured variational inference algorithm is developed for SLDS, a globally convergent DBN inference scheme, and compared with standard SLDS inference techniques to indicate the significant potential of the SLDS approach.
Tracking through singularities and discontinuities by random sampling
80 Citations1999J. Deutscher, Ben North +2 more
Random simulation (particle filtering or Condensation) proves to provide a robust alternative algorithm for tracking that can also deal with these difficult conditions of markerless tracking.
Unsupervised Learning of Word Boundary with Description Length Gain
74 Citations1999Chunyu Kit, Yorick Wilks
A framework for modeling the appearance of 3D articulated figures
68 Citations2002Hedvig Sidenbladh, Fernando De la Torre +1 more
A framework for constructing a linear subspace model of image appearance for complex articulated 3D figures such as humans and other animals and results of tracking single limbs and walking humans are presented.
British Journal of PsychologyAN ALGORITHM FOR THE SEGMENTATION OF AN ARTIFICIAL LANGUAGE ANALOGUE
59 Citations1975J. Gerard Wolff
The mechanism and properties of a computer program which can perform an analogous segmentation of letter strings are described and it seems to model other aspects of cognition besides perceptual segmentation: the use of redundancies to effect economies in storage and retrieval of information, induction, and the importance of context in recognition.
Learning Multi-Class Dynamics
49 Citations1998Andrew Blake, Ben North +1 more
It is shown here how EM can be combined with the CONDENSATION algorithm, which is based on propagation of random sample-sets, and plausible dynamical models are found to emerge from the learning process.